Convective Rainfall Rate


Table of contents

1. Goal of CRR product  
2. CRR algorithm summary description  
3. List of inputs for CRR  
4. Description of CRR outputs 
5. Example of CRR visualisation 

Access to "Algorithm Theoretical Basis Document for the Precipitation Product Processors of the NWC/GEO" for a more detailed description of the algorithm.


Information on the impact of the different corrections, the possible limitations and usefulness of the product can be found in the "CRR Characteristics web"



1. Goal of CRR product

The objective of the CRR product is to estimate the precipitation rate associated to convective clouds. This product provides to forecasters complementary information to other NWC SAF products related to rain and convection monitoring as Precipitating clouds and Cloud type.


2. CRR algorithm summary description

The algorithm developed for the NWC/GEO CRR product assume that clouds being both high and with large vertical extent are more likely to be raining, so that R = f(IR,VIS), being R the rainfall intensity expressed  in mm/h. By other side, the IR-WV brightness temperature difference is a useful parameter for extracting deep convective clouds with heavy rainfall (Kurino, T., 1996).

The basic CRR mm/h value for each pixel is obtained from calibration functions. Calibration analytical functions are generated by combining SEVIRI and Radar data. Composite radar data are compared pixel by pixel with geographically matched MSG data in the same resolution, and the rainfall rate RR is obtained, as a function of two or three variables (IR brightness temperature, IR-WV brightness temperature differences and normalised VIS reflectances):


   RR = f (IR, IR-WV, VIS), for 3-V calibration

   RR = f (IR, IR-WV), for 2-V calibration


The calibration method, based on Rainsat techniques, tries to establish a statistical relationship between VIS reflectances, IR and WV temperatures and the rainfall rates derived from radar data. In summary, a composite radar data were compared pixel by pixel with a geographically matched MSG data in the same resolution and the total rain rate were calculated as a function of the two or three variables (IR brightness temperatures, IR-WV brightness temperature differences and normalised VIS reflectances). The radar data are used only for training the system and are not used directly as part of the output product.

In a second phase, a filtering process is performed in order to eliminate stratiform rain data which are not associated with convective clouds: the obtained basic CRR data are set to zero if all the nearest pixels in a grid of selected semisize (def. value: 3pix) centred on the pixel do not have an equal or higher value than a selected threshold (def. value: 3mm/h). The size of the grid and the filter threshold can be modified by the user through the configuration model.

To take into account the temporal and spatial variability of the cloud tops, the amount of moisture available to produce rain and the influence of orographic effects on the precipitation distribution, several correction factors can be applied to the basic CRR value by the users. So that, the possible correction factors are the moisture correction, the cloud top growth/decaying rates or evolution correction, the cloud top temperature gradient correction  (Gilberto et all, 1998) and the orographic correction (Gilberto et all, 1999).

Moisture correction factor

This factor has been defined as the product of Precipitable Water, PW, in the layer from the surface to 500 hPa and the Relative Humidity, RH, (mean values between the surface and the 500 hPa level) data, obtained from a numerical model. The PWRH factor takes values from 0.0 to 2.0, and the environment is considered dry if PWRH is significantly lower than 1.0 and quite moist if PWRH is greater than 1.0.

Cloud growth rate correction factor

Convective rain is assumed to be associated with growing clouds exhibiting overshooting tops. Consecutive satellite IR images are used to indicate vertically growing and decaying cloud systems.

The cloud growth correction factor, also designated as evolution correction factor, only changes the magnitude of the rain rate through a coefficient if the analysed pixel becomes warmer in the second image.

Cloud-top temperature gradient correction factor

The cloud growth rate correction factor can not be applied when consecutive images are not available. In this case the alternative method of Cloud-top Temperature Gradient Correction is applied.

This correction factor, also designated as gradient correction factor, is based on a search of the highest (coldest) and lowest (less cold) cloud tops. The idea is to search for the pixels that are below the average cloud top surface temperature (local temperature minima) and assume these pixels indicate active convection associated with precipitation beneath.

The hessian of the temperature field is analysed for each pixel with a temperature lower than 250K, in order to search for those pixels with extreme values as is explained in the Algorithm Theoretical Basis Document[enlace]. Different coefficients will be applied modifying the rain rate corresponding to those pixels which have a maximum (meaning that are warmer than its surroundings) and those ones which have neither a local IR temperature maximum nor minimum. Otherwise rain rate is not modified.

Parallax correction

To apply the orographic correction factor is necessary to know the exact cloud position with respect to the ground below. This is not a problem when a cloud is located directly below the satellite; however, as one looks away from the sub-satellite point, the cloud top appears to be farther away from the satellite than the cloud base. This effect increases as you get closer to the limb and as clouds get higher.

When the Parallax Correction is working, a spatial shift is applied to every pixel with precipitation according to the basic CRR value.

Orographic correction

Rainfall amounts are dependent on the atmospheric flow over the mountains and on the characteristics of the flow disturbances created by the mountains themselves.

This correction factor uses the interaction between the wind vector (corresponding to 850 hPa level from the NWP) and the local terrain height gradient in the wind direction to create a multiplier that enhances or diminishes the previous rainfall estimate, as appropriate.

Lightning algorithm

As lightning activity is related with convection, this information has been added to the product as an optional input. Only Cloud-to-Ground lightning flashes are used by this algorithm.

To incorporate this information into the product a rain rate has been assigned to every lightning depending on:

  • the time distance (delta tau) between the lightning event and scanning time of the processing region centre.
  • the location of the lightning
  • the spatial density of lightning in a time interval

Once the precipitation pattern has been computed, it is compared to the CRR precipitation pattern in order to obtain the final product. This final product contains the highest rain rate of the two.


3. List of inputs for CRR 

Dynamic inputs

- Satellite imagery:

The following SEVIRI brightness temperatures and visible reflectances and are needed at full IR spatial resolution   











The SEVIRI channels are input by the user in HRIT format and extracted on the desired region by SAFNWC software package.

* If TPrev10.8µm is not available, the Cloud Growth Rate Correction Factor can not be computed but the Cloud-top Temperature Gradient Correction Factor is computed instead as an alternative. 


- Numerical model:

This information is mandatory for moisture and orographic corrections. When this information is not available, CRR is computed without applying these two corrections.

Parallax correction can run without the NWP parameters using the climatological profile.

For moisture correction:

  • Relative Humidity at 1000, 925, 850, 700 and 500 hPa
  • Dew Point temperature at 2 m
  • Temperature at 2 m
  • Temperature at 1000, 925, 850, 700, 500 hPa
  • Surface Pressure

 For parallax correction:

  • Temperature at 1000, 925, 850, 700, 500, 400, 300, 250 and 200 hPa
  • Geopotential at 1000, 925, 850, 700, 500, 400, 300, 250 and 200 hPa

For orographic correction:

  • U and V wind components in 850 hPa


- Lightning information file for CRR:

A file with information on every lightning (the time of occurrence and the location) occurred in a time interval is mandatory to choose the option of combining the CRR precipitation pattern with the lightning information.


Static inputs

- Sun angles associated to GEO imagery

This information is mandatory for normalising the VIS image when the solar channel is used. It is computed by the CRR software itself using the definition of the region and the satellite characteristics.


- Ancillary data sets:

  • Saturation Vapour table is mandatory for Humidity correction.
  • Saturation Vapour Polynomial Coefficients table is mandatory for Humidity correction.
  • Climatological profile is mandatory for Parallax correction.
  • Elevation mask is mandatory for orographic correction.

- Model configuration file for CRR:

The CRR model configuration file contains configurable system parameters in the product generation process related with algorithm thresholds, numerical model data, corrections to be applied, etc.


4. Description of CRR main outputs

CRR product is coded in NetCDF format and its content is the following:

CRR clases

The rainfall rates obtained by the CRR algorithm expressed in mm/h are converted into twelve classes as it is shown bellow:





rate < 0.2


0.2 = rate < 1


1 = rate < 2


2 = rate < 3


3 = rate < 5


5 = rate < 7


7 = rate < 10


10 = rate < 15


15 = rate < 20


20 = rate < 30


30 = rate < 50


rate >= 50


CRR hourly accumulations     

Rainfall rates from the images in the last hour are used in order to compute the hourly accumulations. This output provides precipitation accumulations from 0.0 to 51.0 mm with a step of 0.2 mm.

CRR intensity in mm/h 

Rainfall rates in mm/h are necessary to calculate the hourly accumulations. This output provides rainfall rates from 0.0 to 51.0 mm/h with a step of 0.2 mm/h.


5. Example of CRR visualisation

Instantaneous Rates

Below is shown an image corresponding to CRR classes output. It has been obtained at full resolution and all corrections have been applied.

Figure 1. CRR classes output corresponding to 9th June 2015 at 12:00Z.


Hourly Accumulations

Below is shown an image corresponding to CRR hourly accumulations output. It has been obtained at full resolution and all corrections have been applied.

Figure 2. CRR hourly accumulations output corresponding to 9th June 2015 at 12:00Z



-         Kidder, S.Q., and T.H. Vonder Haar, 1995: Satellite Meteorology: An Introduction. Academic Press

-         Scofield, R.A., 1987: The NESDIS operational convective precipitation estimation technique, Mon. Wea. Rev., Vol.115, pp.1773-1792.

-         Vicente, G.A. and R.A. Scofield, 1996: Experimental GOES-8/9 derived rainfall estimates for flash flood and hydrological applications, Proc. 1996 Meteorological Scientific User's Conference, Vienna, Austria, EUM P19, pp.89-96.

-         Kurino, T., 1996: A Rainfall Estimation with the GMS-5 Infrared Split-Window and Water Vapour Measurements, Tech. Rep., Meteorological Satellite Centre, Japan Meteorological Agency.

-         Schmetz J., S. S. Tjemkes, M. Gube and L. van de Berg, 1997: Monitoring deep convection and convective overshooting with METEOSAT. Adv. Space Res., Vol. 19, pp433-441.

-         Vicente, G.A., Scofield, R.A. and Menzel W.P. 1998: The Operational GOES Infrared Rainfall Estimation Technique, Bull. American Meteorological Society, Vol. 79, No. 9, pp. 1883-1898.

-         Vicente, G.A., Davenport, J.C. and Scofield, R.A., 1999: The role of orographic and parallax corrections on real time high resolution satellite rainfall estimation, Proc. 1999 Eumetsat Meteorological Satellite Data User's Conferences, EUM P26, pp. 161-168.

-         Grose AME, Smith EA, Chung HS, Ou ML, Sohn BJ, Turk FJ, 2002: Possibilities and limitations for quantitative precipitation forecasts using nowcasting methods with infrared geosynchronous satellite imagery. J. Appl. Meteor., Vol. 41, pp. 763-785.

-         Jorge Sánchez-Sesma and Marco Antonio Sosa: EPPrePMex, A Real-time Rainfall Estimation System Based on GOES-IR Satellite Imagery. IPWG, October 2004, Monterey, California, USA.

-         Bellon, A., Lovejoy, S and Austin, J. L., 1980: Combining satellite and radar data for the short range forecasting of precipitation. Mon. Wea. Rev., Vol. 108, pp.1554-1566.

-    Tapia, A., Smith, J. A., Dixon, M., 1998: Estimation of Convective Rainfall from Lightning Observations, Bull. American Meteorological Society, Vol. 37, pp. 1497-1509.

-     Lábó, E., Putsay, M., Kocsis, Z. and Szenyán, I. 2009: Cross-verification of the Rapid Developing Thunderstorm and the precipitation products of the Nowcasting and Very Short-Range Forecasting SAF. Help Desk VS Reports.

-         Algorithm Theoretical Basis Document for ”Convective Rainfall Rate” (CRR - PGE05 v3.1), 2010

-         Product User Manual for “Convective Rainfall Rate” (CRR - PGE05 v3.1), 2010

-         Validation Report for “Convective Rainfall Rate” (CRR - PGE05 v1.0), 2010